ht-aws-asg-action-plugin

AWS Autoscaling Hollowtrees action plugin

This is an action plugin for the Hollowtrees project. Auto Scaling Groups helps to ensure that the correct number of Amazon EC2 instances are available to handle the load for the application. It can specify the minimum and maximum number of instances in each Auto Scaling group, and Auto Scaling ensures that the group never goes below or above this size. However, when working with spot instances ASG have limitations and these are handled by the plugin.

The plugin interacts with AWS auto scaling groups by changing instances to new ones with better cost or stability characteristics.

When started it is listening on a gRPC interface and accepts Hollowtrees events.

Quick start

Building the project is as simple as running a go build command. The result is a statically linked executable binary.

go build .

Configuration

The following options can be configured when starting the action plugin. Configuration is done through a plugin-config.toml file that can be placed in conf/ or near the binary:

[log]
    format = "text"
    level = "debug"

[plugin]
    port = "8888"
    region = "eu-west-1"

The project is using the standard aws go client library, so AWS credentials can be provided through env variables, instance profiles or config files in the ~/.aws directory.

To run:

./ht-aws-asg-action-plugin

Event types that the plugin can understand:

prometheus.server.alert.SpotTerminationNotice - detaches the AWS instance from the auto scaling group that will be terminated, and starts a new instance with the same characteristics but with a different instance type and spot bid price that will be attached to the auto scaling group instead.

prometheus.server.alert.SpotInstanceTooExpensive - same as above but terminates the instance after detaching it from the auto-scaling group. Can be used to change "expensive" instances to other instance types in an auto scaling group.

Default Hollowtrees plugins

There are a few default Hollowtrees action plugins available:

• Kubernetes action plugin to execute k8s operations (e.g. graceful drain)
• AWS autoscaling group plugin to replace instances with a better cost or stability characteristics